Congress considering HPCC Act renewal

(The following is an edited statement made by Edward D. Lazowska at an October hearing of the House Science Subcommittee on Basic Research on the High-Performance Computing and Communications Act. Lazowska is chair of the University of Washington Department of Computer Science and Engineering and chair of the Computing Research Association's Government Affairs Committee.)

The High-Performance Computing and Communications (HPCC) Act is now about 4 years old and will expire in less than a year. It is an appropriate time to ask what the program has achieved and where we should go from here. As a research society, CRA has closely followed the HPCC Act from its original inception through its implementation as a program.

Overview

Before commenting in detail on the HPCC Act, I would like to make two general points.

1. CRA strongly endorses the findings of the National Research Council's report Evolving the High-Performance Computing and Communications Initiative to Support the Nation's Information Infrastructure. In particular, we would emphasize the following:

• There has been mind-boggling progress in information technology, which pervades most aspects of our lives and our economy.
• The nation that leads in information technology enjoys enormous competitive advantages.
• America owns this leadership today, thanks to a successful and complex interplay between government, academia and industry in support of research. The track record is crystal clear.
• The government's role is also clear. Industry can afford to look ahead only a few years. But as a nation, we must invest for the long term. This fundamental basic research takes place in universities, with government support.
• The HPCC program is a major success. In particular:

a) The emphasis on "high-performance" is appropriate. Cutting-edge information technology is a window on the future. Today's high-performance technology will be the home, office or schoolroom computer of 10 years from now.

b) The emphasis on parallel computing is appropriate. Although there is much more to be done, parallel computing's viability is clear. Many important problems throughout science and engineering have been tackled, and most vendors have some sort of multiprocessor offering.

c) Interagency coordination and cooperation is working exceedingly well. More than any other scientific field, computing research cuts across many agencies. There is no clear lead agency in computing. HPCC represented a bold attempt to create, in a sense, a virtual agency. It is a mechanism for coordinating the programs of diverse agencies, which serve diverse missions but which all have some interest in advancing the state of computing and communications technology. From our perspective outside of government, this experiment is providing greater coherence and direction to the various programs with which we work.

2. CRA believes that continued authorization of the HPCC program--continued funding in these critical research areas and continued strengthening of the interagency process--is essential to the nation for the following reasons:

• The HPCC program is the nation's research and education program in information technology. It is a coordinated, multiagency initiative that supports nearly all our nation's fundamental research and graduate education in information technology. HPCC is much more than the support of supercomputer centers, although the centers have grown into multidimensional institutions that make a variety of contributions to science and engineering. HPCC is much more than research on the highest-performance machines, although these systems are indeed "time machines" that offer an invaluable window into the future. Instead, HPCC is--and always has been--systems, software, networking, human resources and research on information infrastructure technology and applications.
• The interagency coordination fostered by the HPCC Act has been a model success in coordinating the efforts of multiple agencies with multiple approaches. This close coordination can be seen in several specific program initiatives, including the supercomputer centers, the gigabit test beds and the digital library initiative.
• The HPCC program has proven to be appropriately flexible and adaptable. Fundamental research is inherently unpredictable. The program has adapted and evolved in many ways, including an increased focus on software and high-speed digital communications and the addition of the Information Infrastructure Technology and Applications element to the program structure.
• A strategic plan for the future exists. The cooperation and coordination stimulated by the HPCC program has been extended to a strategic planning effort on the part of the National Science and Technology Council's Committee on Information and Communications. The plan, America in the Age of Information, identified six strategic areas "to focus fundamental information and communications research and to accelerate development in ways that are responsive to NSTC's overarching goals, agency mission goals and our nation's long-term economic and defense needs."

The strategic focus areas are global-scale information infrastructure technologies, high-performance/scalable systems, high-confidence systems, virtual environments, user-centered interfaces and tools, and human resources and education. CRA believes such multiagency planning and program implementation efforts are excellent.

* The role of universities is critical. Federally supported university research has played a key role in essentially every aspect of modern information technology: timesharing, computer networking (the Internet), high-power workstations, computer graphics, database technology, very large scale integrated circuit design, reduced instruction-set computing processor architectures, input/output systems based on redundant arrays of inexpensive disks and parallel computing, to name a few.

Universities look to the future. The HPCC program has been a huge success in allowing them to push the frontiers of their research further into the future. It is important to emphasize that university research carried out under HPCC avoids picking winners and losers. The purpose of publicly funded research in high-technology fields is to advance basic knowledge and create new opportunities that, in the medium and long term, industry exploits.

The HPCC Act

The act was organized around four principal sections: systems, software, the National Research and Education Network (NREN), and basic research and human resources. Since that time, the program has been expanded to a fifth component, information infrastructure technology and applications, which emphasizes research on leading-edge applications of information systems in areas of high potential impact, such as education, libraries and public health as well as the technological infrastructure to support them.

1. Systems. When the HPCC program was first being formulated in Congress and the White House, it focused primarily on the largest high-end systems, known as supercomputers. But even at the time the HPCC Act was enacted, the program's focus was broadening to a wider range of architectural goals. The emphasis was not so much on bigger, faster systems based on traditional architectures, but on exploring new, experimental architectures. In particular, researchers were exploring basic questions about the viability of the highly parallel, scalable computer systems and examining different architectural concepts that seemed to have merit. Now we have a better idea of the most promising architectural lines.

Significant challenges remain, particularly the one of how to scale parallel systems to higher numbers of processors. Performance at the chip level is improving at 50% per year, meaning that the performance potential far outweighs our knowledge of how to assemble chips together into productive systems and manage the flow of work through them.

2. Software. The focus at the time was on "grand challenge" computational science. In the intervening years, progress has been made in many computationally demanding areas of basic scientific research.

Over the last four years, however, our vision has broadened substantially. First, as we have progressed in our understanding of the design of scalable parallel architectures, there is an ever greater need to progress on our fundamental understanding of software and algorithms. Although important hardware challenges remain, a proportionally greater emphasis is needed on software. If these advanced architectures will likely be the basis of future everyday desktop systems, research undertaken now on software and algorithms for these leading-edge systems will build the foundation for using them efficiently and effectively in the next century.

Although computational science "grand challenges" remain exciting and important to explore, we now are looking at a wider range of national challenges, applications crucial to the evolution of the nation's information infrastructure.

3. NREN. Five years ago, the Internet was still an academic research network used by university researchers and students. Even then, however, the base of users was broadening to undergraduate schools, libraries, K-12 education and civic networks--so-called "freenets." At the time it was far from clear that NREN would ever be much more than such a specialized system.

The National Science Foundation now has nearly completed the process of spinning the Internet off to the private sector. The Internet continues to grow at an explosive rate. The media frequently run stories about the Internet and the World Wide Web. Commercial firms are fighting each other in the courts over network domain names. Packet-switching communications technology is an important component of the communications service and hardware industry.

NSF, in its concern for the health of US science, needs to ensure that as the Internet becomes commercialized, the needs of researchers and students for specialized advanced data communications services are met. A major research responsibility also remains. As fast as researchers find ways to increase the speed of the networks, increased traffic and demands on new applications consume resources. Thus, there remains an ongoing research agenda in extremely high-speed, large-scale data networking, an agenda that should remain in the next-generation act. Such an agenda would include the following:

• Scaling. For all its growth, the Internet is still relatively small compared with an information infrastructure that would serve all of society. Problems of scaling become even more complex as applications, such as the World Wide Web, are developed that gobble up increasing amounts of communications resources.
• Quality of service. The Internet originally was designed and built as a research network. To be fully usable in a commercial and public arena, its performance must be made more predictable and reliable.
• Security. If valuable assets or sensitive private information are to be transferred over the network, it must be made more secure, and ways to protect information in an insecure environment need to be developed.
• Accounting. Although this may seem to be a simple administrative matter, we have only begun to explore how to measure, manage and account for the flow of information through extremely high-speed networks. Without new insights into how to accomplish this task with minimal overhead, administrative processes could more than double the cost of data transmissions, simply to allow for accounting.
• Technical and legal issues. Many policy and legal issues are being raised as the infrastructure moves more broadly into commercial use, issues such as intellectual property protection, privacy, access to government information, and the First Amendment and censorship. Many of these issues have at least in part a technological component. Careful research could help by identifying possible technological solutions or by clarifying the technological nature of the problem.

4. Basic research and human resources. When the original bill was being considered, CRA repeatedly pointed out the need for a focus on basic research as the necessary underpinning for any HPCC program. So strongly did we feel about this that we insisted that basic research needed to be specifically identified in a fourth section of the bill, even though the other sections arguably could be read to include fundamental investigations. Happily, Congress agreed with us.

We think the need for strong support of basic research is unchanged. If anything, the focus is even more on the need for such fundamental work.

5. Information infrastructure technology and applications. This area of interest has emerged as a fifth program element. I'll elaborate briefly on just two examples of national challenges: educational technology and digital libraries.

Interest is growing in educational technology. The House Science Committee recently held joint hearings with the Senate Labor and Human Resources Committee to explore the potential of information technology to transform education. In those hearings, computer scientists, including Seymour Papert, drew an ambitious and futuristic vision of how information technology could fundamentally transform learning. There is no question that information systems in the future will have the potential to play an enormous role in education. Nor do we doubt that a lot of fundamental computing research needs to be pursued before we can tap that potential.

Similarly, in the area of digital libraries, we face a significant basic research agenda. For all the excitement that rightfully attended the evolution of the World Wide Web, information on the Web has been likened to taking all the books in a large library and dumping them at random on the floor. We don't know how to organize and search for information in such a massive distributed environment. We don't know how best to display it, to use it, to protect intellectual property rights for proprietary data while maintaining access to public information or to protect the privacy of users.

Recommendations

1. We believe that a reauthorization of the HPCC Act would be an important statement by Congress of the need to continue the nation's long-standing commitment to fundamental research in the computer field. It would serve to set general priorities within the computer field and to provide a foundation for interagency coordination. CRA would support such a bill.

2. The focus of the research program needs to be broad but concentrated on fundamental research on the design and use of advanced, leading-edge parallel, scalable computer systems; extremely high-speed data communications; and the connections between the two. In particular, there is an important basic research agenda that underlies so-called national challenges, which will be a basic framework for the nation's infrastructure--education, library, health, government services and so on.

3. The authorization needs to be flexible, allowing the program to adapt quickly as new research targets of opportunity appear.

We understand that money is tight and that there are many claims on it. We also understand the enormous pressures Congress finds itself in when trying to preserve research funding. But we are beginning to realize the enormous return on 50 years of investment in computing research. The information industry is still the fastest-growing--and now biggest--sector of our economy.

We cannot let our investments in the basic research that underpins this field falter. If we fail to invest in research in information technology today, we will lose our leadership tomorrow. And once lost, it would be difficult, if not impossible, to recapture.